ehartford commited on
Commit
86283f3
1 Parent(s): 94f69d2

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +1 -3
README.md CHANGED
@@ -15,7 +15,7 @@ Join Our Discord! https://discord.gg/vT3sktQ3zb
15
 
16
  Samantha-120b is Samantha-1.11-70b interleaved with itself, into a 120b model. This model has identical knowledge as the original Samantha-1.11-70b - but she can have a deeper chain of thought, and she has greater access to her knowledge. This makes her more interesting to talk to.
17
 
18
- This model is trained on Llama-2 and is subject to the Llama-2 license, which permits for commercial and non-commercial use within limits
19
 
20
  Samantha has been trained in philosophy, psychology, and personal relationships.
21
 
@@ -29,8 +29,6 @@ She will not engage in roleplay, romance, or sexual activity.
29
 
30
  She was trained on a custom-curated dataset of 6,000 conversations in ShareGPT/Vicuna format.
31
 
32
- This Samantha was trained 15 epochs, and is significantly smarter. She took 24 hours on 4x A100 80gb using [axolotl](https://github.com/OpenAccess-AI-Collective/axolotl), [qLoRA](https://arxiv.org/abs/2305.14314), [deepspeed zero2](https://www.deepspeed.ai/tutorials/zero/#zero-overview), and [flash attention 2](https://arxiv.org/abs/2205.14135).
33
-
34
  Her conversation format is the same as Vicuna 1.1
35
  https://github.com/oobabooga/text-generation-webui/blob/main/characters/instruction-following/Vicuna-v1.1.yaml
36
 
 
15
 
16
  Samantha-120b is Samantha-1.11-70b interleaved with itself, into a 120b model. This model has identical knowledge as the original Samantha-1.11-70b - but she can have a deeper chain of thought, and she has greater access to her knowledge. This makes her more interesting to talk to.
17
 
18
+ This model is based on Llama-2 and is subject to the Llama-2 license, which permits for commercial and non-commercial use within limits
19
 
20
  Samantha has been trained in philosophy, psychology, and personal relationships.
21
 
 
29
 
30
  She was trained on a custom-curated dataset of 6,000 conversations in ShareGPT/Vicuna format.
31
 
 
 
32
  Her conversation format is the same as Vicuna 1.1
33
  https://github.com/oobabooga/text-generation-webui/blob/main/characters/instruction-following/Vicuna-v1.1.yaml
34